Exportin 1 Inhibition Induces Nerve Growth Factor Receptor Expression to Inhibit the NF-κB Pathway in Preclinical Models of Pediatric High-Grade Glioma

Pediatric high-grade glioma (HGG) is the leading cause of cancer-related death among children (1). Mean survival following diagnosis is 9 to 15 months (2). In diffuse midline glioma (DMG), a subtype of HGG, surgery is usually not feasible due to the diffuse nature of the tumors and, even when resection is a therapeutic option, the combination of surgery and radiation are only transiently effective. To date, no pharmacologic agent has increased survival in DMG, despite more than 200 clinical trials testing available agents (2). In non-DMG HGG, chemotherapy can be effective but is rarely curative and can result in serious and long-lasting morbidity. Effective therapies with less significant side effects than those currently available are an urgent need for HGG.

Nerve growth factor receptor (NGFR), known formally as “Protein tumor necrosis factor receptor superfamily member 16” and also called p75 neurotrophin receptor, is a cell surface receptor that lacks inherent kinase activity but appears to regulate cell growth, survival, differentiation, and proliferation (3). Its effects are context specific, promoting survival or proliferation in some circumstances and inhibiting them in others. NGFR binds all of the neurotrophins, including NGF, brain-derived neurotrophic factor (BDNF), and neurotrophins-3 and -4 (4). In in vitro models of prostate, bladder, and colorectal cancer, NGFR suppresses tumor cell proliferation by regulating progression through the cell cycle, suggesting a tumor suppressor role (5–7). In breast cancer, elevated levels of NGFR are significantly associated with longer disease-free survival and overall survival (8).

Selinexor is an orally bioavailable, reversible small-molecule inhibitor of the karyopherin exportin 1 (XPO1; ref. 9). It is the subject of several phase I to III clinical trials in adult solid tumor and hematopoietic cancers and is also being evaluated in phase I pediatric trials, including one focused on HGG (10). We previously showed selinexor is effective at inducing apoptosis in in vitro and in vivo models of HGG but found tumors eventually grew in vivo leading to animal death following an initial positive response (11), presumably due to the development of resistance. XPO1 mediates the nuclear export of several tumor suppressors as well as numerous other proteins and mRNAs that may be involved in oncogenic pathways (12). Upregulated nuclear export through overexpression of XPO1 is observed in a number of cancers, including HGG, and can contribute to depletion of tumor suppressors such as p53 and other XPO1 cargo molecules from the nucleus (9). By inhibiting nuclear export, selinexor may preserve nuclear levels of tumor suppressors that inhibit cancer cell proliferation. In a human-derived osteosarcoma cell line, selinexor inhibits the prosurvival and proinflammatory transcriptional programs of NF-κB. Selinexor blocks phosphorylation of serine 536 (S536) of the p65 subunit of NF-κB; inhibits phosphorylation of IκB-α, protecting it from degradation; and inhibits the nuclear export of IκB-α, enabling it to bind nuclear NF-κB to inhibit gene transcription (9). Utilization of proteasome inhibition to further preserve cellular IκB-α levels is synergistic with selinexor in inducing tumor cell death (9).

We observed in in vitro models of HGG that selinexor induces NGFR expression, prompting our investigation of the role that NGFR plays in selinexor-induced cell death. Because NGFR interacts with the NF-κB pathway, we hypothesized that selinexor's mechanism of growth inhibition depends at least in part on NGFR-mediated regulation of NF-κB transcriptional activity (4). Our objectives were to identify molecular and phenotypic effects of modulating NGFR expression, including changes in the NF-κB pathway, proliferation rate, and the ability to sustain anchorage-independent growth; to determine whether selinexor treatment recapitulated those changes through induction of increased NGFR levels; and whether NGFR knockdown (KD) results in resistance to selinexor-mediated cell killing. The acquired resistance inherent in the use of small-molecule inhibitors prompted us to also perform drug screening of selinexor in combination with chemotherapeutic agents, including proteasome inhibitors, to identify potentially synergistic combinations for further preclinical investigation.

We designed our study to investigate the mechanism of action of selinexor in HGG using cell culture and orthographic xenograft models; specifically, we sought to determine the role in selinexor's mechanism of action of induced NGFR expression and the extent to which NGFR expression alters the NF-κB pathway.

Primary human pediatric DMG/diffuse intrinsic pontine glioma (DIPG) cell lines derived at autopsy or biopsy were cultured in serum-free medium containing FGF, EGF, and PDGFa/b growth factors in neurosphere (suspension) or adherent conditions. The cell lines utilized include BT245 (derived from a thalamic DMG at biopsy), DIPG4, DIPG6, and DIPG7 (derived from DIPG tumors at autopsy), GBM1 (derived from a pediatric GBM at biopsy), and SF7761 (derived from a DIPG tumor at biopsy; Supplementary Table S1). The identity of all lines was validated by microsatellite DNA profiling and Mycoplasma testing performed prior to and during this project. Additional details regarding methods are contained in the Supplementary Materials and Methods.

Selinexor (Karyopharm Therapeutics; Supplementary Fig. S1A) was dissolved in DMSO (Sigma) to a concentration of 10 mmol/L and stored in accordance with manufacturer's instructions. Selinexor was diluted with PBS or medium to its final concentration and a DMSO level of 0.1% by volume for all experiments. DMSO at 0.1% by volume was used as the vehicle control for in vitro experiments involving drug treatment. In experiments involving measurements of cell viability following treatment by selinexor, three replicates per condition were utilized. Viability assays were 120 hours in length. At the conclusion of each experiment, cell viability was evaluated using CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega) according to manufacturer's instructions. IC₅₀ values were calculated using Prism 7 (GraphPad) or CompuSyn as noted in the text.

Cells to be evaluated for apoptosis levels were plated in triplicate. Caspase Glo 3/7 (Promega) was prepared and added to cells as per manufacturer's instructions. The mixture was incubated for 30 to 60 minutes and then imaged for luminescence.

Cells were treated with DMSO (0.1%) or selinexor (1 μmol/L) for 16 hours, exposed to 50 ng/mL TNFα (to activate NF-κB transcription) for 4 hours, and then immediately lysed to extract protein. Transcriptionally active NF-κB was measured using the Chemiluminescent Transcription Factor Assay Kit (Thermo Fisher Scientific, catalog no. 89859) according to the manufacturer's recommendations.

We used a functional genomics approach, transducing SF7761, DIPG6, and DIPG4 cells through lentiviral delivery of shRNA to target NGFR mRNA for destruction, resulting in stable KD of NGFR protein expression. Two NGFR shRNAs and two control constructs were used in KD experiments. Transduced cells were maintained in puromycin for the duration of the study. Clonal NGFR KD and scrambled shRNA control cell lines were prepared from cells grown in neurosphere dilution assays described below.

We transfected cells with an NGFR ORF plasmid (EX-AO100-M68; GeneCopoeia) containing only the protein-coding open reading frame, or an empty vector control. Transfection efficiency was confirmed by qPCR. Cells were allowed at least 48 hours to begin expressing transfected DNA before being subjected to experimental conditions.

For cell proliferation assays, cells were counted at approximately weekly intervals in Trypan blue using a TC20 Automated Cell Counter (Bio-Rad). Counts were verified periodically using a hemacytometer. Neurosphere dilution (self-renewal) assays were performed by plating 20 or 30 replicates of 1 or 10 cells per well. Cells were grown until neurospheres were established in at least some wells, at which point the wells with neurospheres were counted and the size of the neurospheres measured. For the methycellulose (colony formation) assay, cells were plated in triplicate in a 1:1 mixture of 2× medium and methylcellulose. When visible colonies began to form, the cells were stained with nitrotetrazolium blue chloride and imaged. Counting was performed in ImageJ by converting images to 16-bit black and white, inverting the image, thresholding at 3%, and counting as colonies images comprising 16 or more pixels.

Cells were plated in BioCoat chamber slides (Corning). Unless otherwise noted, in experiments involving selinexor, cells were treated for 16 hours at five times the cell-specific IC₅₀ concentration established by a 120 hours viability assay as described above. We found through time course experiments using varying levels of selinexor treatment that these conditions optimized expression levels for the proteins of interest. Cells to be stained were fixed in formaldehyde, permeabilized, and blocked in 4% BSA. Cells were incubated in primary antibody, rinsed, and then incubated in secondary antibody (Alexa Fluor 488 or 555; Life Technologies). Confocal imaging was performed at 400× using 405 nmol/L (DAPI), 488 nmol/L (Alexa Fluor 488), and 561 nmol/L (Alexa Fluor 555) lasers. Primary antibodies and concentrations were as follows: NGFR (Santa Cruz Biotechnology, sc-8317, 1:200), IκB-α (Santa Cruz Biotechnology, sc-371, 1:200), NF-κB (p65; Cell Signaling Technology, #6956, 1:400), phosphorylated-NF-κB (S536; Cell Signaling Technology, #3033, 1:200). Nuclear expression was quantified in ImageJ by creating a threshold mask in the DAPI channel and using the analyze particles function to select regions of interest (ROI) consisting of cell nuclei. The ROIs were then applied to the corresponding antibody-stained channels and quantified using the individual measure function eliminating particles occupying less than 200 pixels. Mean values for treated samples were computed and evaluated for statistical significance versus control using Student t test (P < 0.05).

Histologic sections prepared from brains of mice implanted with orthographic xenografts of BT-145, an adult patient-derived glioblastoma cell line, were utilized for immunofluorescence staining. Tumor samples were obtained at the death of the mouse from the tumor-related effects or euthanasia. Slides were deparaffinized and antigen retrieval performed. Cells were permeabilized and stained as described above. Confocal imaging and image analysis were performed as above.

We isolated nuclear and cytoplasmic fractions of cell populations utilized. Twenty micrograms of protein per lane was loaded on SDS-PAGE gels and electrophoresis performed followed by wet transfer to polyvinylidene difluoride. We immunoblotted using the antibodies also utilized for immunofluorescence staining, along with actin (cytoplasmic fraction) and lamin A-C (nuclear fraction) loading controls. Blots were imaged using horseradish peroxidase–conjugated secondary antibodies in a G:Box imaging system (Syngene).

We extracted RNA using the RNeasy Mini Kit, and prepared cDNA using the Quantitect Reverse Transcription Kit (both Qiagen). The qPCR assays were conducted on a LightCycler 480 II Instrument (Roche Life Science) using LightCycler 480 SYBR Green I Master Mix (Roche) and KiCqStart SYBR Green Primers (Sigma).

RNA was extracted from approximately 1 million cells per condition following vehicle (0.1% DMSO) or selinexor treatment for 16 hours at five times the cell-specific IC₅₀ concentration. Library preparation and sequencing were performed at the University of Colorado Genomics and Sequencing Core Facility using single-pass 125-bp reads (1 × 125) with approximately 50 million reads per sample. The resulting data were mapped to the human genome (hg38), expression [fragments per kilobase per million mapped reads (FPKM)] was derived, and differential expression analyzed with ANOVA (13–15). Files output for analysis contained read-depth data (FPKM) for approximately 66,000 transcripts. Relative gene expression levels versus control were computed for each transcript in log₂ format, and the data were analyzed using gene set enrichment analysis (GSEA) and specific NF-κB gene sets (Table 1; ref. 16). GSEA results were evaluated using the normalized enrichment score (NES), in which increased expression results in a positive NES and reduced expression in a negative NES. Scores were considered potentially informative from a statistical perspective if the FDR was less than 0.25 (16). Raw and processed RNA-sequencing (RNA-Seq) files are available in the GEO repository at accession number GSE137702.

IκB-α super repressor, an altered form of IκB-α with the mutations S32A and S36A, is resistant to phosphorylation and thus acts as a constitutively active inhibitor of NF-κB (17). We transfected IκB-α super repressor or its empty vector control using the procedure described above for transfection of NGFR cDNA.

For the initial drug screen, we utilized the Approved Oncology Drugs Set (NCI, AOD6) comprising 129 drugs, supplemented by selinexor and AZD2014 (Supplementary Table S2). Cells were plated at 10,000 (DIPG4) or 30,000 (SF7761) cells per well in 100 μL medium in a 96-well cell culture plate and treated with the drugs being screened at a concentration of 1 μmol/L and 0.1% DMSO for 5 days (Supplementary Table S2). DMSO (0.1%) was utilized as a control. The screen for synergy with selinexor was also conducted using the screen drugs at 1 μmol/L and selinexor at its approximate IC₅₀ concentration of 100 nmol/L in DIPG4 cells and 150 nmol/L in SF7761 cells. Cell viability was assayed following 5 days of treatment using the procedure described above in the section titled “Selinexor preparation and treatment.”

Validation testing for synergy between selinexor and candidate drugs identified from the initial drug screen [aldoxorubicin (CytRx), axitinib (Selleckchem), bortezomib (Selleckchem), dactinomycin (Sigma), and vinorelbine (Selleckchem)] was carried out using a viability assay procedure similar to that used for the drug screen. Prior to validation testing, we determined the IC₅₀ of each of the validation candidates by treating cells with a range of drug concentrations. Results were analyzed in Prism 7 and CompuSyn. In the validation testing, we utilized constant ratios of candidate drug and selinexor at the calculated combined IC₅₀ of the two drugs and four half-log concentrations above and below the combined IC₅₀. In each experiment, five concentrations of selinexor alone and the candidate drug alone were also used (IC₅₀ and two half-log concentrations above and below IC₅₀). Results were analyzed using Prism 7 (GraphPad) and the Chou–Talalay method as implemented in the CompuSyn software (18, 19).

We began by examining the effect of selinexor on NGFR mRNA and protein expression in four DMG (BT245, DIPG4, DIPG7, and SF7761) and one non-DMG HGG (GBM1) cell lines in neurosphere culture. We observed increased levels of total cellular NGFR gene expression, as measured by mRNA levels (Fig. 1A). The effect was dose dependent, and the increase ranged from 2- to 24-fold in the HGG cell lines tested. Notably, in BT245 and DIPG7 cells, the NGFR expression level decreased at the largest concentration (1,000 nmol/L). This dose level exceeds the IC₅₀ by 10-fold or more in these cell lines (Supplementary Table S1). We interpret the decrease in NGFR expression at the largest dose level as suggesting interruption of cellular processes and the inception of selinexor-induced cell death. Total NGFR protein expression following selinexor treatment also increased in three of five cell lines, but the effect was less pronounced than the gene expression results (Fig. 1B). To determine the extent to which NGFR induction occurred in a manner suggesting greater cell surface receptor expression, we performed immunofluorescence staining for NGFR without first permeabilizing the treated cells. We found the increased NGFR levels attributable to selinexor treatment occurred predominantly intracellularly in two cell lines (DIPG4 and SF7761) and predominantly at the cell surface in BT245, DIPG7, and GBM1 cells (Fig. 1B and C; Supplementary Fig. S1B). For the DMG cell lines (BT245, DIPG4, DIPG7, and SF7761), sensitivity to selinexor, as measured by the IC₅₀ concentration, correlated closely with the extent to which NGFR expression increased on the plasma membrane as the result of selinexor treatment (R² = 0.977, Fig. 1D). The non-DMG HGG cell line (GBM1) experienced an increase in surface NGFR with selinexor treatment and is highly sensitive to selinexor (IC₅₀ = 61.4 nmol/L) but was not included in the analysis of DMG cells because of expected differences in tumor-related biology between the non-DMG and DMG cell lines.

Because selinexor upregulates NGFR in HGG cell lines, we expected it also to affect the NF-κB pathway. Accordingly, we investigated nuclear and cytoplasmic levels of total NF-κB, IκB-α, and NF-κB phosphorylated at serine 536 [p-NF-κB (S536)] through immunofluorescence staining following selinexor treatment (Fig. 1E–I; Supplementary Fig. S1C). We found by qPCR and RNA-Seq that IκB-α gene expression increased by varying amounts in four of the five cell lines tested (all except SF7761 cells) following selinexor treatment (Fig. 1E; Table 1A). We anticipated selinexor might decrease nuclear NF-κB through increased cytoplasmic sequestration as the result of preservation of IκB-α levels, or decrease free NF-κB, defined as the ratio of nuclear NF-κB to nuclear IκB-α proteins, by increasing nuclear levels of IκB-α. We observed this result only in DIPG7 cells (Fig. 1F and G). We found that selinexor treatment did not significantly alter NF-κB (RELA) gene expression levels except in GBM1 cells. Selinexor, however, produced a consistent decrease in the ratio of p-NF-κB (S536) to total NF-κB (Fig. 1H and I; Supplementary Table S3), implying that in HGG cells selinexor alters the NF-κB transcriptional program through changes in phosphorylation as opposed to changes in nuclear levels of free NF-κB or IκB-α. To further validate the immunofluorescence results, we performed Western blot analyses on nuclear fractionated samples of BT245 and DIPG4 cells following selinexor treatment. We treated with selinexor for 16 hours at two dosage levels (IC₅₀ and 10× IC₅₀ for each cell line). The results confirm a decrease in the phosphorylated-NF-κB/NF-κB ratio with selinexor treatment similar to that seen in the immunofluorescence data (Supplementary Fig. S1D).

Previously we showed that selinexor treatment decreased tumor growth by a factor of 2 (P < 0.001) and prolonged survival by approximately 50% (P = 0.001) in an orthotopic patient-derived xenograft (PDX) mouse model of HGG (11). Tumor samples from these mice showed increased NGFR protein levels following selinexor treatment and, in contrast to the in vitro data, increased IκB-α levels and decreased free NF-κB levels (Fig. 2; Supplementary Fig. S2). When we costained tumor samples for p-NF-κB and NF-κB; however, we did not observe a significant difference in the ratio of p-NF-κB/NF-κB with selinexor treatment (Supplementary Fig. S2D). The different time frames for the in vitro and in vivo analyses may explain the discrepancy. The in vitro quantifications of p-NF-κB/NF-κB were performed after a brief treatment duration (16 hours) designed to prevent the inception of adaptation to the selinexor-induced nuclear export inhibition. The in vivo samples were taken after the tumors had grown to the point that the mice bearing them died or were euthanized, meaning that selinexor's activity had been overcome by resistance mechanisms that resulted in tumor progression.

To explore the functional consequences of selinexor treatment on the NF-κB pathway, we conducted RNA-Seq in four cell lines (BT245, DIPG4, DIPG7, and GBM1) on samples treated with selinexor at five times the cell line-specific IC₅₀ concentration for 16 hours, the same treatment protocol as was used to prepare Fig. 1. The RNA-Seq results confirm that selinexor produces increased expression of NGFR, consistent with our qPCR and immunofluorescence results (Fig. 1A and B), and XPO1 (Table 1A). Increases in XPO1 gene expression with selinexor treatment appear to be a cellular feedback mechanism in response to XPO1 inhibition (20). In three of the four cell lines, selinexor did not produce a significant change in the expression of RELA, which codes for the principal NF-κB (p65) protein of the canonical NF-κB pathway (Table 1A). With regard to the regulators of NF-κB nuclear translocation, the RNA-Seq data show generally increased expression of NFKBIA, which codes for the canonical NF-κB inhibitor, IκB-α, and decreased expression of IKBKG, which codes for IκBγ (NEMO), the regulatory subunit of the kinase that can free canonical NF-κB of IκB-α and enable its nuclear translocation (21). With regard to RELB, the principal NF-κB protein of the noncanonical NF-κB pathway, we found increased expression of RELB binding partner p100 (encoded by NFKB2) but decreased expression of IKKα (encoded by CHUK), the regulatory kinase required to release p100 from RELB to enable its nuclear translocation (22).

We also performed gene set enrichment analysis (GSEA) using several NF-κB–related gene sets (Table 1B). GSEA showed increased TNF and NGFR (three of four cell lines) signaling via the NF-κB pathway (Table 1B). TNFα and NGFR (because it includes TRAF signaling domains) are members of the tumor necrosis receptor superfamily (23). With regard to overall NF-κB transcriptional activity, both negative and positive regulation were identified (Table 1B). Consistent with the decreased expression of IKKα, GSEA identified decreased signaling in the noncanonical RELB NF-κB pathway (GO_NIK_NF_KAPPAB_SIGNALING) in three of four cell lines. Together, the RNA-seq data on individual NF-κB pathway genes and GSEA results suggest that selinexor treatment: (i) induces TNF family (NGFR) signaling via NF-κB; (ii) inhibits the nuclear import of NF-κB (p65) through the canonical pathway by increasing IκB-α and decreasing IκBγ (NEMO) expression; (iii) exerts context-dependent up- and downregulation of NF-κB transcriptional activity; and (iv) inhibits signaling through the noncanonical (RELB) NF-κB pathway.

By varying NGFR gene expression levels, we further explored whether selinexor's mechanism of action depends upon its induction of NGFR expression. We used lentiviral delivery of shRNA to target NGFR mRNA for destruction in two DMG cell lines, SF7761 and DIPG6. We chose SF7761 because it responds well to transduction and, like most DIPG cell lines, has the H3K27M mutation (Supplementary Table S1). We added the DIPG6 cell line because it is also H3K27M mutant and is in addition p53 null (Supplementary Table S1). Because selinexor has been observed to preserve p53 in the nucleus (p53 is an XPO1 cargo; ref. 24), we decided to also test NGFR KD in a line that lacks p53 regulation. NGFR KD at levels of 90% or more increased ratios of NF-κB/IκB-α mRNA and induced greater levels of free NF-κB compared with control cells (Fig. 3A–C; Supplementary Fig. S3A). NGFR KD cells proliferated more quickly than control cells (Fig. 3D; Supplementary Fig. S3B and S3C). They also showed greater capacity for self-renewal than control cells, forming neurospheres in nonadherent growth conditions and colonies in methylcellulose (Fig. 4A–C). NGFR KD cells plated at a density of one cell per well formed neurospheres in 12 of 20 test wells versus one of 20 for control cells (P < 0.001, Fig. 4A). The NGFR KD neurospheres were significantly larger than the control neurosphere (0.132 mm vs. 0.07 mm diameter, P < 0.05; Fig. 4B). In SF7761 cells, although both NGFR KD and control cells formed neurospheres in all wells, the neurospheres formed by NGFR KD cells were significantly larger than those formed by control cells (0.359 mm vs. 0.26 6 mm diameter, P < 0.001, Fig. 4B; Supplementary Fig. S4A). When grown in methylcellulose, SF7761 NGFR KD cells formed colonies more readily than control cells (P = 0.03); in DIPG6, the NGFR KD cells had a nonstatistically significant increase in colony formation (P = 0.1, Fig. 4C; Supplementary Fig. S4B). To investigate the potential drivers of the increased self-renewal and proliferation observed in NGFR KD cells, we quantified the mRNA and protein levels of several cell lineage markers. Cells with NGFR KD had increased mRNA levels of GFAP, MAP2, synaptophysin (SYP), and TUBB3, and decreased OLIG2, whereas cells overexpressing NGFR (through cDNA) had decreased GFAP expression (Fig. 4D and E). At the protein level, NGFR KD cells had increased GFAP and SYP expression, and cells overexpressing NGFR had decreased GFAP, Nestin, and SYP expression (Fig. 4D and E). These results suggest that increased NGFR expression may drive a cell toward a differentiated state that might disrupt proliferation of undifferentiated glial tumor cells.

We treated NGFR KD cells with selinexor to determine whether lowered NGFR expression rendered them resistant to selinexor's effects. NGFR KD cells showed little or no increase in NGFR expression when treated with selinexor compared with a large induction of NGFR expression observed in control cells (Fig. 5A). NGFR KD cells were resistant to selinexor-induced cell death, having an IC₅₀ of 133 nmol/L, whereas control cells had an IC₅₀ for selinexor of 21.9 nmol/L (P < 0.05, Fig. 5B). Apoptosis levels based upon caspase 3/7 activation in NGFR KD cells following selinexor treatment were approximately 63% of control at 24 hours after treatment (P < 0.004) and 53% of control at 48 hours after treatment (P < 0.02, Fig. 5C).

A previous report in the literature described decreased tumor cell proliferation through downregulation of the NF-κB pathway as the result of selinexor treatment (9), so we measured levels of NF-κB and its inhibitor, IκB-α, in NGFR KD cells treated with selinexor. We found that selinexor had little effect on IκB-α mRNA or protein levels in NGFR KD cells when compared with vehicle (Supplementary Fig. S5). Free NF-κB levels, however, increased versus control in NGFR KD cells treated with selinexor (P = 0.008, Fig. 5D). In cells overexpressing NGFR, mRNA expression of IκB-α increased whereas NF-κB expression remained unchanged in DIPG4 cells and IκB-α expression remained unchanged whereas NF-κB expression decreased in SF7761 cells (Fig. 5E). The net result was that in cells overexpressing NGFR, selinexor treatment decreased expression of free NF-κB by a statistically significant amount in SF7761 cells and by an amount that was just short of statistical significance (P = 0.09) in DIPG4 cells (Fig. 5E).

To confirm whether the insensitivity of NGFR KD cells to selinexor resulted from de-repression of the NF-κB pathway, we transfected cells with IκB-α super-repressor, a plasmid coding for a variant of IκB-α (S32A, S36A) that resists phosphorylation and is therefore constitutively active in repressing the transcriptional activities of NF-κB. We anticipated the IκB-α variant would sensitize wild-type cells to selinexor by increasing the pool of IκB-α available to inhibit NF-κB activity when confined to the nucleus by selinexor. Our results confirmed that cells transfected with the IκB-α variant were more sensitive to selinexor than wild-type cells (Fig. 5F). Transfection of NGFR KD cells with the IκB-α variant restored sensitivity to selinexor when compared with cells transfected with a scrambled shRNA control (Fig. 5G), suggesting that NGFR KD mediates cellular survival (and possibly selinexor resistance) via the NF-κB pathway.

In our orthotopic PDX model of HGG, selinexor used as a monotherapy initially slowed tumor growth, but resumption of tumor outgrowth eventually occurred (11). Accordingly, to combat resistance, we performed an in vitro screen in two HGG cell lines (DIPG4 and SF7761) pairing selinexor with each agent from a panel of FDA-approved anticancer drugs to identify potentially synergistic combinations. We identified several candidates for further evaluation based upon efficacy versus vehicle and a suggestion of potential synergy with selinexor (Fig. 6A). The candidates for which we undertook further validation of the initial screening results included axitinib (a multi-tyrosine kinase inhibitor), aldoxorubicin (a DNA intercalator substituted for other anthracyclines because of its lipophilicity, allowing brain penetration), bortezomib (a proteasome inhibitor), dactinomycin (a DNA synthesis inhibitor), and vinorelbine (a microtubule inhibitor that causes M-phase arrest). In initial experiments to validate the candidate drugs, axitinib and aldoxorubicin failed to show any synergistic effect. Further in vitro investigations demonstrated synergistic effects between selinexor and both bortezomib and dactinomycin in both DIPG4 and SF7761 cells; vinorelbine was synergistic with selinexor in DIPG4, but not SF7761, cells (Table 2; Supplementary Fig. S6). The proteasome inhibitor bortezomib was active at sub-10 nmol/L levels and had combination index values of 0.37 and 0.13, respectively, in DIPG4 and SF7761 cells when combined with selinexor, suggesting the potential of a useful synergistic interaction between nuclear export inhibition and proteasome inhibition in slowing HGG tumor cell growth (Table 2).

NGFR expression inhibits proliferation and induces apoptosis in several contexts, but has the opposite effect in others. NGFR overexpression induces apoptosis in primary cortical neurons, pheochromocytoma cells (PC12 cell line), and adult glioblastoma (GBM) cell lines U343 and U373 (25). As noted above, NGFR expression suppresses tumor cell proliferation in in vitro models of prostate, bladder, and colorectal cancer and is associated with significantly longer disease-free survival and overall survival of patients with breast cancer (5–8). In invasive cells isolated from U87 and U251 adult GBM cell lines; however, increased levels of NGF enhance invasiveness in vitro, and overexpression of NGFR increases invasiveness and migration in vivo (26). Similarly, in brain tumor–initiating cells (BTIC) derived from three adult GBMs and one giant cell GBM, NGFR KD leads to decreased cell viability and decreased proliferation, based on Ki67 expression (27). In several human cancer-derived cell lines, as well as a mouse xenograft model, NGFR inactivates p53 through MDM2-dependent and -independent mechanisms (28).

Our results align with studies in which NGFR acts as a tumor suppressor. Knocking down NGFR expression in HGG increased cell proliferation rates and the self-renewal ability of HGG cells. NGFR KD increased the protein expression of glial cell marker GFAP and reduced levels of the neuronal markers SYP and TUBB3, whereas overexpression of NGFR decreased GFAP and increased levels of SYP and TUBB3. GFAP expression correlates with higher histopathologic grade in HGG tumors (29). Conversely, increasing TUBB3 expression through a pharmacologic agent decreases proliferation and significantly increases survival in mice in a HGG xenograft model (30). Our phenotypic and molecular data suggest that increasing NGFR expression in HGG cells may induce terminal differentiation that slows proliferation.

NGFR KD also leads to changes to NF-κB pathway members that suggest NGFR's antiproliferative effect is transmitted at least in part through the NF-κB pathway. NGFR KD decreased nuclear levels of IκB-α, but not NF-κB, resulting in an increase in the ratio of NF-κB/IκB-α in the cell nucleus, suggesting that transcriptionally active NF-κB (which can induce proliferative activity) increases as a result of NGFR KD. Although NGFR has been characterized as an activator of NF-κB, this action occurs in response to ligand binding (4). Our results indicate that, in contrast, expression of the receptor in the absence of ligand suppresses NF-κB, while silencing NGFR promotes activation. Because the receptor binds intracellular signaling components, such as RIP2 and TRAF6, which mediate NF-κB signaling (31–33), it is possible that increasing NGFR expression limits the availability of these factors, thereby reducing the basal signal.

Additional aspects of our results support the inference that modulation of NF-κB's transcriptional activity is important in driving the cellular proliferation that accompanies NGFR KD. Our GSEA analysis identified an NGFR-mediated transcriptional signal and transcriptional changes (negative and positive) in NF-κB transcriptional activity as the result of selinexor treatment. GSEA also showed selinexor produces a decrease in noncanonical NF-κB signaling. HGG cells with stable NGFR KD treated with selinexor-resisted NGFR induction, cell death, and apoptosis as compared with vehicle-treated cells. In the NGFR KD cells, levels of free NF-κB increased with selinexor treatment while they were unchanged in cells stably transduced with nontargeted shRNA. Transfection of NGFR KD cells with a constitutively active form of IκB-α that resists dissociation from NF-κB restored selinexor sensitivity to levels approximating that of cells with normal NGFR levels. The restoration of selinexor sensitivity as the result of NF-κB inhibition is likely to be clinically significant. The possibility that independently targeting an aspect of the NF-κB pathway could enable a reduced selinexor dose level to have equivalent effect, or an equivalent selinexor dose to have a potentiated effect, has the potential to reduce toxicity and increase efficacy, especially in the treatment of brain tumors, in which drug concentration in the central nervous system is critical. In total, our investigation of NGFR KD in HGG cells suggests that upregulation of transcriptionally active nuclear NF-κB levels resulting from NGFR KD can drive proliferation and death resistance.

In our in vitro experiments, the extent to which selinexor induced the expression of NGFR on the surface of DMG cells correlated with sensitivity to selinexor. In two of the DMG cell lines (DIPG4 and SF7761), however, increased NGFR levels were predominantly intracellular. It is possible that the increased intracellular levels were a manifestation of ongoing transport to the plasma membrane, but also possible that intracellular NGFR plays some role in selinexor's mechanism. Interestingly, selinexor treatment produced a greater increase in NGFR protein expression than would be suggested by the results of our NGFR overexpression experiments (compare Fig. 1A–C with Fig. 5E). This result suggests that selinexor's mechanism of inducing NGFR expression is not simply mediated by an increase in NGFR transcription.

Selinexor treatment tended to induce an increase in IκB-α expression (4/5 cell lines) but also tended to increase NF-κB expression (4/5 cell lines), with the result that selinexor treatment, in contrast to NGFR induction, tended to increase the ratio of nuclear NF-κB/IκB-α. Our RNA-Seq data suggested that GBM1 cells had several differences in NF-κB pathway gene expression as compared with the DMG cell lines. GBM1 cells showed increases in both RELA and NFKB2 (p100) expression and no decrease in CHUK (IKKα) with selinexor treatment, suggesting that selinexor did not downregulate the noncanonical NF-κB pathway in the GBM cells as it did in the DMG cell lines. Together, the data show some differences in response to selinexor in terms of NGFR induction and NF-κB pathway response in our in vitro experiments, which are areas for further exploration.

Selinexor treatment, however, produced a consistent response across all cell lines in decreasing nuclear levels of p-NF-κB (S536) as a proportion of total NF-κB. Selinexor also decreased absolute levels of nuclear p-NF-κB (S536) in three of the five cell lines tested. Phosphorylation of NF-κB at S536 has proliferative effects in HeLa cells and increases invasiveness and motility of prostate cancer cell lines (34, 35). As compared with wild-type NF-κB, the transcriptional targets of p-NF-κB (S536) are enriched in genes that promote inflammatory and oncogenic processes (35, 36). p-NF-κB (S536), however, can also induce apoptosis in breast and colon cancer cell lines under certain conditions (37). On the basis of our results, p-NF-κB (S536) appears to play an oncogenic role in pediatric HGG. We note that selinexor was effective at low concentrations (approximately 100 nmol/L) in cell lines that experienced little or no induction of cell surface NGFR, implying that selinexor works through pathways in addition to NGFR to impact p-NF-κB (S536) expression and may also work through mechanisms apart from NF-κB. This is expected given the many different signaling pathways and cellular processes that XPO1-mediated nuclear export influences. Our ongoing work in HGG includes further investigation of additional pathways by which selinexor exerts its mechanism of action.

In addition to elucidating the significance of NGFR in selinexor's mechanism of action, we also sought to identify potentially synergistic chemotherapeutic agents to pair with selinexor to overcome acquired resistance. In in vitro experiments, selinexor was synergistic with the proteasome inhibitor bortezomib, a finding that is consistent with results reported in other cancer models (38, 39). Preclinical investigations combining selinexor and proteasome inhibition in multiple myeloma models showed synergistic interactions that included inhibition of NF-κB signaling (40). A clinical trial in relapsed or refractory multiple myeloma showed high response rates to a combination of selinexor, bortezomib, and dexamethasone (40). Proteasome inhibition can lead to cell death by preventing the degradation of IKB-α, resulting in repression of prosurvival transcriptional activity of NF-κB (41). In other cases, however, proteasome inhibition activates NF-κB, but nonetheless induces cell death, possibly through transcriptional activity of NF-κB that induces proapoptotic gene expression (42). The synergistic relationship between selinexor and proteasome inhibitors that we observed may be the result of two different mechanisms of NF-κB regulation (NGFR signaling and proteasome inhibition), a possibility that we are studying further with marizomib, a lipophilic proteasome inhibitor that penetrates the blood–brain barrier (43).

Our results suggest NGFR expression plays a tumor-suppressing role in HGG and that selinexor's mechanism of action in HGG cell lines depends at least in part upon induction of NGFR expression that inhibits the proliferative or antiapoptotic transcriptional program of p-NF-κB (S536). In vivo results in an orthotopic xenograft model of HGG are consistent with the conclusion that selinexor's induction of NGFR expression downregulates antiapoptotic aspects of the NF-κB pathway. Because of selinexor's consistent induction of NGFR expression in in vitro and in vivo models of HGG, NGFR might serve as a pharmacodynamic marker of selinexor activity. Finally, our drug combination results suggest that combining selinexor with a drug that is independently directed at the NF-κB pathway, such as a proteasome inhibitor that inhibits degradation of IKB-α, may be an effective method to increase the therapeutic value of selinexor and overcome acquired resistance.

B.D. Carter reports receiving a commercial research grant from Karyopharm Therapeutics. Y. Landesman is a VP of Research and Translational Development at Karyopharm Therapeutics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.A. DeSisto, P. Flannery, T. Kashyap, R. Vibhakar, A.L. Green

Development of methodology: J.A. DeSisto, A. Pathak, N.J. Bales, S. Venkataraman, B.D. Carter, A.L. Green

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.A. DeSisto, R. Lemma, N. Philips, E. Moroze, Y. Landesman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.A. DeSisto, R. Lemma, S. Mestnik, N. Philips, N.J. Bales, T. Kashyap, A.L. Kung, Y. Landesman, A.L. Green

Writing, review, and/or revision of the manuscript: J.A. DeSisto, A. Pathak, T. Kashyap, A.L. Kung, B.D. Carter, R. Vibhakar, R. Vibhakar, A.L. Green

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.A. DeSisto, S. Mestnik, Y. Landesman

Study supervision: R. Vibhakar, A.L. Green

DIPG 4 and DIPG 6 cell lines were kindly provided by Dr. Michelle Monje (Stanford University, Stanford, CA); DIPG7 and GBM1 by Dr. Angel Montero Carcaboso (Hospital Sant Joan de Deu, Barcelona, Spain); SF7761 by Dr. Nalin Gupta (University of California, San Francisco, San Francisco, CA); and BT245 by Dr. Keith Ligon (Dana-Farber Cancer Institute, Boston, MA). The IκB-α super repressor plasmid, pBabe-Puro-IKBalpha-mut (super repressor; Addgene plasmid #15291), and its empty vector control, pBabe-Puro-Myr-Flag-IKBKE (Addgene plasmid #15293), were gifts from William Hahn. Imaging experiments were performed in the University of Colorado Anschutz Medical Campus Advanced Light Microscopy Core, supported in part by Rocky Mountain Neurological Disorders Core Grant Number P30NS048154 and by NIH/NCATS Colorado CTSI (grant no. UL1 TR001082). Antigen retrieval on histology slides was performed by the University of Colorado Denver Research Histology Shared Resource. Analysis of drug combination experiments was performed using CompuSyn (www.combosyn.com). This work was supported by 1K08 NS102532-01, the Luke's Army Pediatric Cancer Research Fund St. Baldrick's Foundation Fellowship, and a TeamConnor Community Impact grant (all to A.L. Green), as well as NS107456 (to B.D. Carter).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.